Lupus & antiphospholipid syndrome
Machine learning: Predicting hospital length of stay in patients admitted for lupus flares
Authors' conclusion
Our results were consistent with the established literature and showed promise in ML over traditional methods of predictive analyses, even with rare rheumatic events such as lupus flare hospitalizations.
Study context
- Population
- Patients admitted for systemic lupus erythematosus (SLE) flares, National Inpatient Sample Database 2016–2018; 90 demographic and comorbidity features
- Participants
- 5,831 patients
- Exposure
- Four machine learning (ML) models for predicting length of stay (LOS)
- Outcome
- Prediction of hospital LOS (accuracy, ROC-AUC, PR-AUC, F1-score)
Methodological detailsmachine learning prediction models, National Inpatient Sample Database 2016-2018, SHAP feature importance
- Machine learning prediction models
- National Inpatient Sample Database 2016-2018
- SHAP feature importance
Main finding
As reported in the abstract
Prediction of hospital LOS (best-performing model)
| Group | ROC-AUC | PR-AUC | F1 score | Accuracy |
|---|---|---|---|---|
| XGB | 0.87 | 0.61 | 0.56 | 95% |
XGB
ROC-AUC0.87
PR-AUC0.61
F1 score0.56
Accuracy95%
Observational data: associations, not effects.
Further findings
Features with most impact on LOS (SHAP)
- Most significant: need for a central line, acute dialysis, acute renal failure
- Other top features: renal and infectious comorbidities
Original publication
Machine learning: Predicting hospital length of stay in patients admitted for lupus flares. Lupus. Published 13 October 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →